Top chip vendors gather this week for the second annual summit of a landmark U.S. research program to drive semiconductors forward.
Researchers from academia, government, and industry are meeting in Detroit this week for the second annual summit for the Electronics Resurgence Initiative (ERI). The ERI program is spending $1.5 billion of U.S. federal funds over five years to drive the semiconductor industry forward at a time when traditional silicon scaling is showing diminishing returns.
It’s still early days for more than a dozen advanced research projects being discussed at the event. Organizers point to the caliber of companies taking part in the program as an indicator of progress.
Intel, GlobalFoundries, Nvidia, Qualcomm, and Xilinx are among active participants in the program. Several have top executives speaking at the event this week.
“There’s a large number of companies that were not in our [previous] programs that are in ERI,” said Mark Rosker, head of the branch of the Defense Advanced Research Projects Agency (DARPA) that oversees ERI. “That didn’t just happen coincidentally; we attracted and involved them.
“What’s unique is [that] we cluster something close to 20 programs that have a concerted strategy, each with its own specific objectives, planned for a single effect.”
Projects span new kinds of circuits, fresh approaches to chip stacking, advanced materials, and new design flows, as well as new techniques for machine learning. “ERI is about driving innovation for a new wave in electronics when geometric scaling no longer makes as much sense,” he said.
Earlier this year, the Semiconductor Industry Association called for triple the level of federal spending on basic research given the sector’s need to find new technology levers of growth. “We’re sensitive to that discussion,” said Rosker.
ERI is one of a handful of efforts, he added, pointing to a separate initiative expected to spend more than $2 billion to bolster the quality and security of the government’s semiconductor supply chain. “Moore’s Law is the background on which this stage is set … but whether Moore’s Law goes on forever or stops tomorrow, we want to run faster than the other guys.”
After adding many new projects last year, Rosker’s goal is to turn ERI into more “of a flow than a flood … over the next three years, we will be doing steady updates and reappraisals. Progress will be more like what DARPA is used to.”
That said, DARPA did launch at least one new program this year. It is focused on accelerators for real-time machine learning. And one day of the summit this week includes open-ended workshops in tech areas that could spawn future projects. Videos and presentations from the event are expected to be posted at the site in about two weeks.
There’s plenty on the horizon these days, including proposals for 6G cellular based on terahertz-band networks.
“I first came to DARPA in 2003, and my first program was on terahertz — for a while, I was the terahertz guy at DARPA,” said Rosker, who developed programs on gallium nitride, terahertz electronics, and quantum cascade lasers. “A lot of my first stint in DARPA was trying to close the terahertz gap, and it’s gratifying to see [that] it’s happening with compound semiconductors, new circuits, amplifiers, receivers, and the like used in systems at 100 GHz and above.”